Estimating Future Costs of CO2 Capture Systems Using Historical ...

Proceedings of GHGT-8, Int’l. Conf. on Greenhouse Gas Control Technologies, Trondheim, Norway, June 2006

Estimating Future Costs of CO2 Capture Systems Using Historical Experience Curves Edward S. Rubin1, Sonia Yeh2, Matt Antes3, Michael Berkenpas1 and John Davison4 1

Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, Pennsylvania, USA 2

Center for Urban and Regional Studies, University of North Carolina, Chapel Hill, North Carolina, USA 3

Energetics, Incorporated, Columbia, Maryland, USA 4

IEA Greenhouse Gas R&D Programme, Stoke Orchard, Cheltenham, UK

Abstract Reductions in the cost of technologies as a result of learning-by-doing, R&D investments and other factors have been observed over many decades. This study uses historical experience curves as the basis for estimating future cost trends in CO2 capture technologies applied to four types of electric power systems: pulverized coal (PC) and natural gas combined cycle (NGCC) plants with postcombustion CO2 capture; coal-based integrated gasification combined cycle (IGCC) plants with precombustion capture; and coal-fired oxyfuel combustion for a new PC plants. We assess the rate of cost reductions achieved by other process technologies in the past, and by analogy with capture plant components estimate future cost reductions that might be achieved by power plants employing CO2 capture. Effects of uncertainties in key parameters on projected cost reductions also are evaluated via sensitivity analysis. Keywords: CO2 capture, cost estimates, experience curves, technology innovation, PC plants, IGCC plants, NGCC plants, oxyfuel plants Introduction Given the growing worldwide interest in CO2 capture and storage (CCS) as a potential option for climate change mitigation, the expected future cost of CCS technologies is of significant interest. Most studies of CO2 capture and storage costs have been based on currently available technology. This approach has the advantage of avoiding subjective judgments of what may or may not happen in the future, or what the cost will be of advanced technologies still in the early stages of development. On the other hand, reliance on cost estimates for current technology has the disadvantage of not taking into account the potential for improvements that can affect the long-term competitiveness of CO2 capture systems in different applications, and the overall role of CCS as a climate mitigation strategy. To address this problem, most large-scale energy-economic models used to assess global climate change mitigation policies and strategies assume some degree of technological improvement over time. While models (and modellers) differ in the approach used to represent technological change (for example, endogenous vs. exogenous rates of improvement), there is currently little empirical data to support assumptions regarding future CO2 capture costs for power plants and other industrial processes. The objective of the present study is to develop projections of future cost trends based on historical observations for other technologies relevant to CO2 capture systems. Study Methodology In this study, we first develop a set of experience curves characterizing historical cost trends for seven 1


technologies relevant to power plants with CO2 capture. These are: flue gas desulfurization (FGD) systems, selective catalytic reduction (SCR) systems, gas turbine combined cycle (GTCC) plants, pulverized coal (PC) boilers, liquefied natural gas (LNG) production plants, oxygen production plants, and steam methane reforming (SMR) systems for hydrogen production. Average learning rates are derived for the capital cost and operating and maintenance (O&M) cost of each technology. To estimate future cost trends for plants with CO2 capture, we first decompose each of the four power plant designs into major process areas or sub-systems that include all equipment needed to carry out certain functions such as power generation, air pollution control, or CO2 capture. We then apply a learning rate to each sub-system based on judgments as to which of the seven case study technologies offers the best analogue to the power plant process area in question. The cost of the total plant is then calculated as the sum of all process area costs for increasing levels of total installed capacity. A classical learning curve is then fitted to the total cost trend to obtain a learning rate for the overall plant with CO2 capture. We also quantify the effect of uncertainties in component learning rates and other key parameters. Although technologies and costs for CO2 transport and storage are outside the scope of this study, these components clearly are critical to a complete CCS system. The following sections provide additional details of the study methodology and results obtained. Case Study Results The experience curves used in this study to characterize cost trends have the form: Υ = ax-b, where, Y is the specific cost of the xth unit, a is the cost of the first unit, and b (b>0) is a parametric constant. The quantity 2–b is defined as the progress ratio (PR). It implies that each doubling of cumulative production or capacity results in a cost savings of (1 – 2–b). The latter quantity is defined as the learning rate (LR). Values of PR and LR are commonly reported as a fraction or percentage for each doubling of cumulative installed capacity or production [1]. Table 1 summarizes the learning rates for capital and O&M cost for the seven technologies examined. Results for three of the technologies (FGD, SCR and GTCC) are based on previous studies [2–4], while the remaining four (Figure 1) are newly derived. Detailed descriptions and discussions of each technology are presented elsewhere [5]. All learning rates derived in this study fall within the range reported in the literature for an array of energy-related technologies [6]. Factors contributing to longterm declines in capital and O&M costs included improvements in technology design, materials, product standardization, system integration or optimization, economies of scale, and reductions in input prices. Table 1. Summary of learning rates for capital and O&M costs, and whether a cost increase was observed during the early stages of commercialization. Learning Rate* Capital Cost O&M Cost Flue gas desulfurization (FGD) 0.11 0.22 Selective catalytic reduction (SCR) 0.12 0.13 Gas turbine combined cycle (GTCC) 0.10 0.06 Pulverized coal (PC) boilers 0.05 0.07-0.30 LNG production 0.14 0.12 Oxygen production 0.10 0.05 Hydrogen production (SMR) 0.27 0.27 Technology

Initial Cost Increase? Yes Yes Yes n/a** Yes n/a n/a

*Fractional reduction in cost for each doubling of total production or capacity. **n/a=not available.

Table 1 also indicates that four of the seven technologies displayed an increase in cost during the early stages of commercialization, as illustrated in Figure 1(a) and Figure 2. Such cost increases relative to pre-commercial estimates often are not reflected in the long-term learning rates reported in the 2


literature. In the context of the current study, the potential for costs to rise before they fall is an important finding affecting projections of future cost trends, as elaborated below. Theoretical liquefaction unit cost

1000

Subcr itica l PC U nit C o st (1 9 9 4 $ /kW )

Liquefaction capital cost ($/tpa)

Actual liquefaction unit cost

100.0 -0.22

y = 269x 2 R = 0.52

1942, EF=29.9% 1965 1999, US DOE EF=37.6% -0.08

y = 515.00x PR = 0.95

100

10.0 0.1

1

10

0

100

1

10

100

1000

Cumulative World Pulverized-Coal Plant Installed Capacity (GW)

Cumulative LNG produced (Mta)

100000

H 2 U nit Price (2 0 0 0 cents per hundred cubic feet)

Real capital cost (US $2003/tpd)

100

-0.157

y = 94254x 2

R = 0.43

10000 1000

10000

-0.47

y = 2174.30x PR = 0.72

10 -0.47

1 1,000

1000

y = 724.77x PR = 0.72

Estimated Capital Charges Estimated non-fuel O&M cost 10,000

100,000

1,0

Estimated Cumulative World Hydrogen Produced by SMR (billion cubic feet)

Cumulative Oxygen Production since 1980 (Billion cubic feet)

Figure 1. Capital cost experience curves derived in this study for four processes: (a) LNG production, (b) PC boilers, (c) oxygen production, and (d) hydrogen production via steam methane reforming. 8

300

1982

1980

250

1982

200

1990 1995

150 1975

100

1974 1972 (1000 MW, eff =80-90%) 1968 (200 MW, eff =87%)

50

O&M Costs (1997$/MWh)

Capital Costs ($/kW) in 1997$

1976

7 1980

6

1990

1976

5

1995 1975

4 3

1974

2

1968 1972

(200 MW, eff =87%) (1000 MW, eff =80-90%)

1

0

0 0

10

20

30

40

50

60

70

80

90 100

0

Cumulative World Wet FGD Installed Capacity (GW)

10

20

30

40

50

60

70

80

90 100

Cumulative World Wet FGD Installed Capacity (GW)

Figure 2. Capital and O&M cost trends for wet limestone FGD systems at a new coal-fired power plant in the U.S. (500 MW, 90% SO2 capture), including cost studies conducted during the period of early commercial applications.

Application to Power Plants with CO2 Capture A number of recent studies have estimated the cost of CO2 capture at power plants [7]. Here we use the Integrated Environmental Control Model (IECM) developed at Carnegie Mellon University [8] to estimate the current cost of four plant types (PC, NGCC, IGCC and oxyfuel). IECM costs are comparable to other reported costs under similar assumptions. In this study, plants are assumed to have 3


a net output of approximately 500 MW, a levelized capacity factor of 75%, and a capture system that removes 90% of the CO2 produced and compresses it to 13.8 MPa. Starting with the estimated current cost, we use the historical learning rates reported in Table 1 to project the future costs of major power plant sub-systems (Table 2) as new plant capacity is built. This approach allows the cost of different plant sections to change at different rates, reflecting differences in the technological maturity of each plant type and sub-system. Component costs are then summed to obtain the total plant cost as a function of total installed capacity. From this, a learning curve of the form Υ = ax-b is derived for the overall plant. One drawback of this approach is that it does not explicitly include potential cost increases that may arise when integrating components that have not yet been proven for the application and/or scale assumed. For example, no IGCC power plant has yet combined CO2 capture with a gas turbine fired by a H2-rich fuel gas at a scale of 500 MW. Nor has an oxyfuel combustion plant compressing concentrated CO2 yet been demonstrated at a commercial scale. Since there is no easy or reliable method to quantify potential cost increases during early commercialization (a common phenomenon also seen in several of the case studies), we instead assume that any such costs effectively delay the onset of learning until later generations of the plant or process are designed, deployed and operated for a period of time. With additional experience, the higher plant costs incurred initially are gradually reduced (via learning-by-doing and continued R&D). The cumulative capacity at which the total plant cost equals the currently estimated cost (Cmin, a parameter of the analysis) is when learning (cost reduction) is assumed to begin. Similarly, a capacity parameter Cmax defines the end point of the projected learning curve. Other variables in the analysis are the current (initial) capacity of each plant sub-system, and a set of multipliers to reflect additional experience from continued deployment of plant components in applications other than power plants with CO2 capture (sometimes called cluster learning). Full details of all assumptions and calculation procedures are described in Ref. [5]. Table 2. Sub-systems for each of the power plants analyzed. NGCC Plant GTCC system CO2 capture system CO2 compression Fuel use

PC Plant Boiler/turbine area Air pollution controls CO2 capture system CO2 compression Fuel use

Oxyfuel Plant Air separation unit Boiler/turbine area Air pollution controls CO2 distillation CO2 compression Fuel use

IGCC Plant Air separation unit Gasifier area S removal/recovery CO2 capture system CO2 compression GTCC system Fuel use

Results for CO2 Capture Plants Table 3 shows the overall learning rates for the total levelized cost of electricity (COE) for each plant from the onset of learning (a variable in the study) to a future point where the total installed capacity of each system reaches 100 GW worldwide. The nominal values reflect a set of base case assumptions for each plant sub-system, while the ranges reflect uncertainty in the component-level learning rates. The nominal learning rates show a 3–5% decrease in COE for each doubling of capacity; with uncertainty the range is 1–8%. Based on these rates, Table 4 shows the overall change in COE. The largest cost reduction (18%) is seen for the IGCC system and the smallest (10%) for the oxyfuel system. The results with learning rate uncertainties show a broader range of COE reductions, from 3–26%. The sensitivity of results to other parameters is shown in Table 5, which displays projected trends in capital cost as well as COE. Overall ranges are slightly greater than those in Table 4. Combustion-based plants, whose cost is dominated by relatively mature components, show generally lower learning rates than gasification-based plants. For similar reasons, the cost of CO2 capture (defined as the cost difference between plants with and without capture at any point in time) declines faster than the total cost of generation (a reduction of 13–40% across the four plant types, as compared to 10–18% for COE) [5]. 4


Table 3. Learning rates for total cost of electricity (excluding transport & storage costs) Technology NGCC Plant PC Plant IGCC Plant Oxyfuel Plant

Learning Rates for Total Plant COE (excl transport/storage) r2 Nominal Range 0.033 1.00 0.006 - 0.048 0.035 0.98 0.015 - 0.054 0.049 0.99 0.021 - 0.075 0.030 0.98 0.012 - 0.049

Table 4. Overall change in cost of electricity after 100 GW of capture plant capacity Technology NGCC Plant PC Plant IGCC Plant Oxyfuel Plant

Cost of Electricity (excl transport/storage) Nominal ($/MWh) Range ($/MWh) Initial Final % Change Range % Change 59.1 49.9 15.5 46.1 - 57.2 3.2 - 22.0 73.4 62.8 14.4 57.8 - 68.8 6.2 - 21.3 62.6 51.5 17.6 46.4 - 57.8 7.7 - 25.8 78.8 71.2 9.7 66.7 - 75.8 3.9 - 15.4

Table 5. Summary of additional sensitivity study results NGCC Sensitivity Case Nominal Base Case Assumptions Learning Starts with First Plant Learning up to 50 GW Current Capture Capacity = 0 GW Non-CSS Exp. Multipliers = 2.0 Natural Gas Price = $6.0/GJ FCF = 11%, CF = 85% PC Sensitivity Case Nominal Base Case Assumptions Learning Starts with First Plant Learning up to 50 GW Current Capture Capacity = 0 GW Non-CSS Exp. Multipliers = 2.0 Coal Price = $1.5/GJ FCF = 11%, CF = 85% IGCC Sensitivity Case Nominal Base Case Assumptions Learning Starts with First Plant Learning up to 50 GW Current Gasifier Capacity = 1 GW Above + H2-GTCC = 0 GW Non-CSS Exp. Multipliers = 2.0 Coal Price = $1.5/GJ FCF = 11%, CF = 85% Oxyfuel Sensitivity Case Nominal Base Case Assumptions Learning Starts with First Plant Learning up to 50 GW Current Boiler Capacity = 0 Non-CSS Exp. Multipliers = 2.0 Coal Price = $1.5/GJ FCF = 11%, CF = 85%

Learning Rate 0.022 0.014 0.018 0.029 0.030 0.022 0.022

Capital Cost ($/kW) Initial Final Learning % Change Rate Value Value 916 817 10.8% 0.033 916 811 11.5% 0.028 916 849 7.3% 0.031 916 786 14.2% 0.037 916 783 14.4% 0.036 925 826 10.7% 0.033 918 820 10.7% 0.034

COE ($/MWh) Initial Final % Change Value Value 59.1 49.9 15.5% 59.1 47.0 20.4% 59.1 52.0 12.0% 59.1 48.8 17.4% 59.1 49.0 17.1% 76.1 64.2 15.7% 51.6 43.3 16.1%

Learning Rate 0.021 0.013 0.018 0.026 0.029 0.021 0.021

Capital Cost ($/kW) Initial Final Learning % Change Value Value Rate 1,962 1,783 9.1% 0.035 1,962 1,764 10.1% 0.024 1,962 1,846 5.9% 0.031 1,962 1,744 11.1% 0.042 1,962 1,723 12.2% 0.068 1,965 1,786 9.1% 0.035 1,963 1,785 9.1% 0.039


Learning Rate 0.050 0.029 0.044 0.057 0.088 0.062 0.050 0.048

Capital Cost ($/kW) Initial Final Learning % Change Value Value Rate 1,831 1,505 17.8% 0.049 1,831 1,448 20.9% 0.032 1,831 1,610 12.1% 0.045 1,831 1,460 20.3% 0.055 1,831 1,285 29.8% 0.078 1,831 1,432 21.8% 0.054 1,834 1,507 17.8% 0.048 1,832 1,516 17.2% 0.047

COE ($/MWh) Initial Final % Change Value Value 62.6 51.5 17.7% 62.6 48.6 22.4% 62.6 54.9 12.2% 62.6 50.2 19.7% 62.6 45.9 26.6% 62.6 49.5 20.8% 68.4 56.6 17.3% 47.2 39.2 16.9%

Learning Rate 0.028 0.013 0.023 0.054 0.038 0.028 0.028

Capital Cost ($/kW) Initial Final Learning % Change Value Value Rate 2,417 2,201 9.0% 0.030 2,417 2,160 10.7% 0.017 2,417 2,291 5.2% 0.025 2,417 2,008 16.9% 0.056 2,417 2,122 12.2% 0.044 2,421 2,204 9.0% 0.030 2,418 2,202 9.0% 0.031


5


Concluding Remarks Projections of technological change are a critical factor in analyses of alternative futures, and the impacts of policy interventions to address issues such as global climate change. In this context, the results of this study can be used to help project and bound estimates of future cost trends for power plants with CO2 capture based on historical rates of change for similar technologies. A study of this nature also has important limitations that must be recognized. For one, while the concept of a constant learning rate is a convenient and widely-used measure to characterize technological change, often it is an over-simplification of actual cost trends for large-scale technologies [9]. For example, several technologies in this study displayed cost increases during early commercialization, followed by subsequent decreases. In other cases, actual cost trends are better represented by an S-shaped curve, in which learning is initially slow, then accelerated, and then gradually slow again [1,9]. Alternative representations of technological learning, including models that account for additional factors such as R&D spending, are a subject of on-going research, and future developments in this area may provide insights beyond the scope of the present study. We also note that this study is based on incremental improvements to existing technologies, which historically has been the dominant mode of technology innovation [10]. However, if radically new CO2 capture technologies were to be developed the resulting cost reductions could be greater than those estimated here. Within the current framework, a more extensive set of sensitivity analyses could provide a more detailed picture of how alternative assumptions influence reported results. Extensions of the current analysis also could incorporate the costs of CO2 transport and storage technologies and their projected trends, as well as improvements in CO2 capture efficiency and its impact on future cost. Such analyses could advance our understanding of potential improvements in the cost-effectiveness of CO2 capture and the cost of CO2 avoided. Software included with the current study [5] can be used to further analyze such options. Acknowledgements Support for this research was provided by the IEA Greenhouse Gas Programme. We also wish to thank Rodney Allam, Jon Gibbins, Howard Herzog, Keywan Riahi, Leo Schrattenholzer and Dale Simbeck for their guidance and review this work. The authors alone, however, remain responsible for its content. References [1] Perspectives on experience, Boston Consulting Group, Inc., Boston, Massachusetts, 1968. [2] Rubin, E. S., S. Yeh, et al., Experience curves for power plant emission control technologies, International Journal of Energy Technology and Policy, 2(1/2): 52-69, 2004 [3] Yeh, S., E. S. Rubin, et al., Technology innovations and the experience curves for NOx control technology, Journal of Air and Waste Management Association, 55 (2):1827-1838, 2005 [4] Colpier, U. C. and D. Cornland, The economics of the combined cycle gas turbine—an experience curve analysis, Energy Policy, 30 (4): 309-316, 2002 [5] Estimating future trends in the cost of CO2 capture technologies, Report No. 2006/5, IEA Greenhouse Gas R&D Programme (IEA GHG), Cheltenham, UK, February 2006. [6] McDonald, A. and L. Schrattenholzer, Learning rates for energy technologies, Energy Policy, 29: 255261, 2001. [7] Special report on carbon dioxide capture and storage, Intergovernmental Panel on Climate Change, Cambridge University Press, 2005 [8] Integrated Environmental Control Model (IECM) version 5.0.2, http://www.iecm-online.com, 2005 [9] Yeh, S., E. S. Rubin, et al., Uncertainties in technology experience curves for integrated assessment models, International Journal of Energy Technology and Policy, 2006 (in press). [10] Alic, J.A., D.S. Mowery and E.S. Rubin, U.S. technology and innovation policies lessons for climate change, Pew Center on Global Climate Change, Arlington, VA, November 2003

6